Project Details
Projekt
Acceleration platform for the simulation of solidification cracks during laser beam welding
Applicants
Dr.-Ing. Arnd Koeppe; Harald Köstler; Dr.-Ing. Michael Selzer
Subject Area
Joining and Separation Technology
Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale
Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale
Term
since 2021
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 434946896
The project aims to develop a digital acceleration platform for multiscale simulation processes in laser beam welding that integrates research data management (RDM) and artificial intelligence (AI). The total duration of the project is 96 months, with phase 2 lasting 48 months. In multiscale modeling and simulation, such as solidification cracks in laser beam welding, there are many scientific processes that benefit from automatization, AI and RDM. Examples include parameter identification, coupling of multiple scales and the development of surrogate models. These processes can be automatized and accelerated in a data-driven environment. In phase 2, a digital acceleration platform integrating RDM and AI will be developed. This platform will enable data-integrated coupling of scales, control, and prediction using AI and rigorous RDM. This will enable simulations with more extensive data volumes, high throughput, and more repetitions. Sub-goals: - Develop a material acceleration interface for laser beam welding processes. - Further development and integration of AI methods for multiscale simulations. - Creation of demonstrators for parameter identification and scale bridging. - Adaptation of workflows, CI pipelines, and support in research data management. Approach: The acceleration platform consists of simulation, RDM, and AI components. The project partners' simulation tools are connected to the Kadi4Mat RDM solution via workflows and continuous integration. In phase 2, the open-source modules KadiAI and CIDS will be used as a starting point to integrate machine learning and AI methods into the Kadi ecosystem. The missing component, the interface for material acceleration, will be developed based on KadiAI and CIDS. Demonstrators: - Iterative identification of simulation parameters and mapping of solidification cracks at the microstructure level. - Data-integrated direct coupling between the scales as a plugin for the FE² multiscale approach. In the long term, a generic acceleration platform, KadiAccelerate, coupled with Kadi4Mat will be developed. This platform can facilitate all research efforts using Kadi4Mat as an RDM tool. The structure remains generic, modular, and decentralized, with plugins and microservices connected to high-performance computing applications. The development of acceleration platforms is challenging and requires significant investment. Thanks to the preliminary work on RDM, AI, workflows, and integration, as well as the existing infrastructure of the Kadi ecosystem, the realization of the acceleration platform in phase 2 is possible.
DFG Programme
Research Units
